The Cutting Edge : PLOWSHARES : ‘Fuzzy Logic’ Professor Turns His Focus to Tumor Detection

Share via

James Keller once used a computer strategy called “fuzzy logic” in a quest to help soldiers find enemy tanks in the jungle. Now he’s using the technique to find tumors in the human body.

Keller, a professor of computer and electrical engineering at the University of Missouri at Columbia, spent 12 years trying to develop fuzzy-logic computer algorithms that would help separate tanks from trees and other objects around the vehicles.

But his work never found its way onto the battlefield because the Cold War ended and his Pentagon funding evaporated.


Now Keller and several colleagues in Missouri are working on a project to apply fuzzy logic to medical imaging, particularly the mammogram X-rays used to detect breast cancer.

The team hopes its techniques will help doctors find potentially cancerous patterns in the images more quickly and more often. Their techniques come into play after the image has been scanned, its contents digitized and fed into a computer for analysis.

“If it found anything suspicious, it would highlight that area and flag it for the radiologist,” Keller said.

Keller’s group also hopes its fuzzy logic can improve the usefulness of biopsies, in which human tissues and fluids are removed to see if they contain abnormal cells. Case in point: The computer could better identify chromosomes that might indicate genetic health defects.

Pathologists looking at chromosomes under a microscope can face difficulties. The chromosomes can be “bent, they’re overlapping, sometimes they’re compressed,” Keller said. “It’s not a real certain situation.”

Fuzzy logic itself isn’t easy to define, and its usefulness has been hotly debated since the concept was spawned by Berkeley engineer Lotfi Zadeh in the mid-1960s. In essence, it’s an effort to get computers to go beyond “yes or no” and “true or false” calculations and to understand shades of truth--that is, to mix what amounts to judgment and intuition with hard facts to solve complex problems.


Keller’s algorithms involve giving the computer--his team uses Sun Microsystems workstations and Apple Macintoshes--substantially more information than in conventional techniques, which could be useful in making a difficult decision.


For example, the computer would attempt not only to determine whether or not a mammogram revealed a small blot, but also whether the blot was round or jagged, whether that meant it was a tumor and, if so, whether such a tumor is usually benign or malignant.

Will Keller’s work find real use this time? He thinks so. His best guess is that a business will license the technology and bring it to market within five years.